Averaging Pitot Tube

What an Averaging Pitot Tube Is and What Is Being Decided

The mechanism explains the name. A classic pitot tube inserted through one wall stops a single column of fluid and reads one stagnation pressure at one point of the velocity profile; on a fully developed turbulent profile that point may sit 15 to 20 percent away from the local velocity the assumption wants, and every elbow upstream moves it. An averaging pitot tube attacks that weakness structurally: the probe spans the pipe, its multiport pitot face samples the profile at several chords at once, and the internal manifold averages those pressures in the hardware before any transmitter sees them. The result is a differential pressure primary element with a signal that survives moderate profile distortion, installed through one weldolet instead of a spool piece.

What is actually being decided in this purchase is a metering point on a large line at minimum installed cost and minimum permanent pressure loss, with an uncertainty number written into the contract. The buyer is not choosing a meter body, because there is none: the pipe is the meter body. The buyer is choosing a probe construction, a mounting family, a port layout, a differential pressure transmitter span, and the density compensation chain, and the RFQ must state all five because the probe alone meters nothing. Vendors of this hardware class include the Rosemount Annubar 485 family and Meriam averaging probes, but every serious manufacturer sells to the same logic: one insertion, one manifold, one transmitter.

The boundary against neighbors can be stated in one breath here and is argued fully in the selection chapter. A single point pitot tube measures one chord location and carries the full profile uncertainty with it; it belongs in traverses, test rigs, and clean circular duct work. An orifice plate flowmeter is the cheap installed standard with heavy permanent loss and a line stop for every plate change. A venturi tube and a flow nozzle buy accuracy with line sized castings and welding. The averaging probe takes the ground between them: accuracy below a calibrated venturi, far above a point reading, at an installed cost that scales with line size barely at all.

The standards position must be stated honestly because datasheets blur it. ISO 5167 covers the geometric DP primaries, orifice plates, nozzles, and venturi tubes, and it does not contain an averaging pitot geometry. Averaging probes are covered by ASME MFC-12M, which treats the multipoint averaging element and its coefficient framework, and in practice by manufacturer calibration data; gas duty inherits the AGA report chain, AGA 3 for orifice runs, AGA 8 detail compressibility in the correcting device, where those apply. An RFQ that asks for "ISO 5167 compliance" on this element is asking for something no manufacturer can supply; the defensible ask is ASME MFC-12M methodology plus a stated calibration basis.

The purchase is a station, not a bar. The probe is sold with a mounting connection, often a packed gland or a flanged tap; the differential reaches the transmitter through a 3 or 5 valve manifold, and on steam the impulse lines need condensate pots and a mounting geometry that keeps the legs filled. Dirty fluids may justify remote seals on the transmitter instead of impulse lines at all. Gas and steam metering need a temperature transmitter and a pressure transmitter feeding density compensation, on modern sites handled in the transmitter or a flow computer. The RFQ should name the whole station so one supplier answers for the uncertainty of the system, because the probe contributes only part of the error budget and none of the compensation.

How Multipoint Averaging Turns Pressure into Flow

The physics is stagnation pressure plus wake pressure. Fluid arriving at the upstream ports is brought to rest locally, so each port reads static pressure plus one half rho V squared at its chord, where rho is density and V the local velocity. The downstream face ports sit in the sheltered wake behind the bar and read a pressure depressed below line static, typically 60 to 80 percent of the upstream dynamic head for common bar shapes. The transmitter sees the difference between the manifold averaged upstream value and the manifold averaged downstream value. Flow then follows the DP relation in words: velocity is proportional to the square root of the differential divided by density, times a dimensionless flow coefficient published by the manufacturer as a function of Reynolds number and probe geometry, and volumetric flow is that velocity times pipe area.

Port placement is where the averaging is engineered. Equal area layouts place ports so each senses an equal annular slice of the pipe cross section; log Tchebycheff layouts place them at the roots of a Chebyshev polynomial, which integrates a polynomial approximation of the profile exactly and is the layout most catalogs default to for round pipe, commonly four opposed wall ports plus a center port. The downstream averaging matters as much: the low side ports span the wake so that bar alignment errors partially cancel. A probe rotated 10 degrees off the flow axis biases the coefficient measurably, which is why installation tolerances in chapter 6 are written in degrees, not shrugged at.

The signal magnitudes decide the transmitter. Averaging probes are low signal devices: at the same line velocity the differential is roughly one fifth to one tenth of what a beta 0.5 orifice plate would produce, so spans of 25 to 250 mbar cover most real duties, with high velocity steam reaching a few hundred mbar. Because differential varies with the square of velocity, a loop running at 30 percent of maximum flow produces only 9 percent of maximum differential, and at 10 percent flow, 1 percent. That square law is the honest reason turndown lands at 4:1 to 10:1 with a smart differential pressure transmitter doing square root extraction: below that floor the signal disappears into transmitter noise, and a 0.075 percent of span accuracy class, typical of a good DP cell, is a floor on what the loop can resolve, not a detail.

Density closes the calculation. Liquids near ambient often ignore it, but steam and gas metering live on compensation: line density is computed from a temperature transmitter and a pressure transmitter input, from steam tables, or from an equation of state, and the square root is taken on differential divided by that density. The 4 to 20 mA loop from each transmitter lands in the DP transmitter math, a flow computer, or the DCS; where the money matters, the compensation runs in a device with certified arithmetic and an audit trail rather than in whatever controller had spare capacity. The flow meter accuracy printed in the brochure assumes the density chain is right; a gas line with a frozen pressure input can be wrong by tens of percent with a healthy probe.

Pressure loss is the argument that sells the element on big lines. The bar obstructs only a few percent of the pipe cross section, and of the small differential it creates, vendors cite recovery ratios around 0.95, meaning roughly 95 percent of the measured differential is recovered downstream, against about 0.6 recovery for a sharp edged orifice at beta 0.5, and the pitot differential itself is a fraction of the orifice one. On a DN 400 steam header at 20 m per second the permanent loss is single digit millibar, a rounding error against pipe friction, while an orifice station on the same line burns recoverable blower or boiler horsepower every hour of its life. When the line is large and the fluid is expensive to push, the loss column in the chapter 4 comparison table is worth more than the accuracy column.

Types, Construction, and Duty Boundaries

Two structural families cover the catalog. The cantilever probe mounts through one wall and hangs across the pipe unsupported, standard through roughly DN 600 and acceptable wherever fluid dynamic loading is modest; it is the cheapest installation because it needs one tap. The dual support, or through, probe enters through a fitting on one side and is supported at a second fitting on the opposite wall, or arrives as an opposed pair, and it is the correct construction for large lines and high velocities where a cantilever bar would shed vortices and fatigue. Within both families the sensing cross section is a machined profile, the Annubar style T bar being the most recognizable, with the upstream face carrying the chordal ports.

Mounting hardware is a separate line on the RFQ and the main driver of installed cost. A packed gland mount welds a socket plus gland to the pipe, the probe passes through a compression packing, and the joint tolerates repositioning; it is the everyday choice for DN 80 to DN 600 lines. A flanged mount bolts a flanged probe into a weld neck flange and suits higher pressure and temperature, including saturated and superheated steam to about the 200 bar g class, vendor dependent, usually with dedicated isolation. The retractable family adds a heavy duty packing extraction assembly with a safety chain or cable so the probe can be pulled under pressure onto a full bore isolation valve, which is how hot tapping and in service inspection happen on lines that cannot stop.

Materials and limits follow the fluid. Wetted parts are 316L as the default, with Hastelloy C-276 specified for wet chloride and sour gas duties where chloride stress cracking rules out standard stainless; packings are PTFE or graphite, graphite holding up at steam temperatures. Size coverage is genuinely wide, DN 80 to DN 3000 plus rectangular duct probes, because nothing scales with the bore except bar length. The limit that bites is velocity: the same cross flow that makes the signal makes von Karman vortex shedding on the bar, and every manufacturer publishes velocity limit tables by line size and fluid density, in steam commonly in the tens of meters per second at mid sizes, below which the probe life is indefinite and above which fatigue cracking of the bar root is a scheduled event, not a surprise.

The duty boundaries deserve a blunt paragraph. Below about DN 80 the probe crowds the pipe, and at the small end a variable area flowmeter already does the job cheaper. Liquids carrying tar, polymer, or heavy scale plug ports and belong on other technologies unless a purge is engineered. Wet gas with liquid slugs hammers a cantilever bar and needs drains, slope, and honesty about what the reading means. And any duty where the money transfers on the meter number should weigh the calibrated accuracy discussion in chapter 4 before assuming the probe is good enough. The table below compresses the mounting decision, which is the one buyers get wrong most often.

Key Specification Parameters

The datasheet rows that decide fitness are few and every one of them should carry a number in the RFQ. The table below lists them with the ranges a competent quotation lands inside; the paragraphs after the table cover the two rows buyers misuse, accuracy and differential span.

Accuracy is the row most often padded. Uncalibrated, an averaging pitot tube carries plus or minus 2 to 5 percent of rate depending on how developed the profile is, and no catalog sentence removes that. Wet calibrated against a reference meter or a prover in situ, the honest claim is about plus or minus 1 percent of rate, sometimes slightly worse, and the calibration certificate is part of the deliverable. This will always lose a mass accuracy contest against a coriolis flowmeter, which reads mass directly at an order of magnitude better uncertainty, and the comparison is not the point: the coriolis is bought per line size and the probe is bought per tap. The RFQ should state which accuracy regime is being purchased and demand the basis, ASME MFC-12M coefficient method with or without in line calibration.

Differential span is the row most often left to the vendor, and it should not be. The RFQ should demand the computed differential at maximum, normal, and minimum flow for the stated density, so the buyer can see where the square law starves the transmitter at low flow and can check that the transmitter span chosen actually covers the duty with margin. The comparison table below places the element against its DP primary neighbors on the numbers that drive selection.

A quotation that is silent on coefficient versus Reynolds number, on velocity limits per line size, or on the computed differential at your flows is not a quotation, it is a price. Demand the coefficient curve, the velocity limit table, material certificates EN 10204 3.1 for the wetted bar, and the calibration basis in writing. Those four documents are what separate a metering point from a metal order.

Selection Decision Factors and RFQ Language

The decision sequence runs as a checklist, and each step closes one escape route a vague quotation could use.

1. State the line size, fluid, pressure, temperature, and the full flow range with densities at each point; the vendor computes differentials from exactly these numbers.
2. Check the velocity limit table for the line size and density; if steam or compressed gas velocity is near the limit, specify dual support and say so.
3. Decide the accuracy regime: internal indication accepts uncalibrated, allocation and energy accounting demand wet calibration and a certificate.
4. Choose the mounting family from chapter 3 against shutdown access, with retractable hardware where the line cannot stop.
5. Fix the transmitter span from the computed differential, and state the manifold, impulse, or remote seal design for the fluid.
6. Define the density compensation chain: none for simple liquids, temperature transmitter and pressure transmitter inputs for steam and gas.
7. Verify straight run availability, 10 to 20 diameters upstream and 5 downstream, or add a flow conditioner to the order.
8. Require the documents: coefficient curve, velocity limits, material certificates, calibration basis, and installation tolerances in degrees.

Against neighboring meter types the honest map runs as follows, and it belongs here rather than in the page title. The single point pitot tube remains a traverse and test instrument; as a fixed installation it carries the whole profile in its error band. The venturi tube and flow nozzle are the accuracy and steam heritage choices, but they are line sized bodies, and above DN 300 their weight and welding dominate the bill. The orifice plate flowmeter is the installed standard of the world, cheap and understood, with permanent loss and plate access as its taxes. A coriolis flowmeter wins outright on mass accuracy and loses on line size economics above roughly DN 150, where the meter cost curve goes vertical. An electromagnetic flowmeter is the full bore answer for conductive liquids, no pressure loss and no probes in the line, where conductivity and a shutdown for installation exist. A clamp on ultrasonic flowmeter suits clean liquids with acoustic coupling and is attractive on water lines that cannot be touched; a multipath ultrasonic flowmeter is the high end gas custody alternative at a different budget. A vortex flowmeter is the clean steam and gas workhorse, line sized, needing turbulent Reynolds and healthy piping. A turbine flowmeter meters clean low viscosity hydrocarbons beautifully and wears with solids. A thermal mass flowmeter reads dry gas mass directly with no density compensation at all and fails quietly on wet gas.

At the small and simple end, a variable area flowmeter still indicates local flows honestly where no power exists, and where the loop needs only proof of flow rather than a measurement, a flow switch is the correct cheap instrument and the averaging probe is the wrong purchase. The probe earns its line on the big, the hot, and the inaccessible: steam headers, air mains, gas laterals, water intake lines, and ducts, with the thermal mass flowmeter as the dry gas alternative and the insertion probe as the only family that hot taps into any of them.

Steam deserves its own paragraph because most probe failures are steam failures. On steam headers the commercial contest is usually between this element and a vortex flowmeter: the probe hot taps with near zero permanent loss, while the line sized meter buys its installation certainty with a shutdown. Saturated and superheated service to the 200 bar g class, vendor dependent, is standard product, but the station must be engineered: dedicated isolation for extraction, dual chamber arrangements on the impulse legs, condensate pots mounted so the legs stay filled, transmitter below the taps on steam, above the taps on gas, insulation on the legs to stop winter split flows, and drains or a vertical mount on wet fuel gas so condensate cannot pool on the ports. The application table below turns the chapter into the form a project file carries.

The RFQ block should be written once and reused: line size and schedule, fluid with chemistry notes, full flow range with densities, the computed differential requested at three flows, probe family and material, mounting family with isolation details, port layout, accuracy regime with calibration basis, transmitter span with manifold and impulse or seal design, compensation inputs required, straight run available with an isometric sketch, and the document list from the checklist. Written that way, every bidder prices the same station and the commercial comparison compares instruments instead of assumptions.

Installation, Standards, Failure Modes, and Maintenance

Installation quality starts with location and orientation. Give the probe 10 to 20 pipe diameters of straight run upstream and 5 downstream, and where close elbows, valves, or expansions make that impossible, either add a flow conditioner or accept the accuracy consequence in writing. The port face must point dead upstream: a 10 degree rotation error biases the coefficient enough to see in the totalizer, so the install procedure should mark the orientation before the packing is tightened, and dual support bars must be aligned so both ends carry the same clock position. On hot taps, the weldolet and valve go in under pressure per procedure, and the probe enters through the full bore valve only after the tap is proven open; the safety chain on retractable hardware is not decoration, it is the last line between the bar and the plant.

The impulse system is half the meter. Steam lines get condensate pots and a differential pressure transmitter mounted below the taps so the legs stay liquid filled; gas lines put the transmitter above so condensate drains back; liquids keep both legs equal and short. A 3 valve manifold is the minimum for zero checks, a 5 valve manifold where the line cannot come out of service for maintenance. Dirty or plugging fluids justify remote chemical seals and capillary, deleting the impulse legs entirely at some cost in temperature rating. Whatever the geometry, the legs are heat traced and insulated where winter can freeze them, because a frozen leg does not stop the flow reading, it corrupts it quietly, which is worse.

The standards frame is short and should be quoted exactly. ASME MFC-12M is the method document for averaging pitot type elements and their coefficients. ISO 5167 governs the other DP primaries and does not cover this geometry; a specification that cites it here is signaling confusion, and the fix is to cite MFC-12M plus the manufacturer calibration basis. Gas duty leans on the AGA chain where it applies, AGA 3 context for orifice metering comparisons and AGA 8 compressibility inside whatever device corrects to base conditions. Material acceptance runs through EN 10204 3.1 certificates for wetted parts, and NACE MR0175 applies where the gas is sour. Calibration, when bought, should be traceable and referenced to the actual installed line size, not a nominal class.

Failure modes arrive in a known order. Port plugging leads on dirty service: tar, polymer fines, and condensate scale find the small chordal holes first, the differential drifts low, and the fix is rodding through the gland, a purge connection engineered at order time, or an honest decision that the fluid is wrong for the element. Vibration fatigue follows on high velocity steam and gas: von Karman shedding loads the bar at its frequency, the vendor velocity tables exist precisely to keep installed duty below the fatigue regime, and a probe found with cracks at the root fitting is an installation that ignored them. Packing leaks show up on packed gland mounts as slow staining; gland adjustment or repack on a schedule beats emergency extraction. Rotation error and profile distortion, the two silent bias sources, are caught by commissioning checks and by re-verifying orientation after nearby piping work. And the transmitter itself drifts, which is why chapter maintenance exists.

Maintenance is cheap and periodic. Zero the transmitter monthly against the equalizing valve; blow down impulse legs on schedule where the fluid allows; inspect and re-grease retractable hardware on the vendor interval; verify probe orientation and gland torque at every nearby turnaround; and recalibrate the transmitter on a bench cycle, with an in line probe verification only when the accuracy regime demands it. Where a downstream flow switch or interlock rides on the same line, proof test it on the same schedule so the whole station is covered in one outage window. File the computed differential, the zero history, and any drift findings: that file is what turns the next RFQ for the next header from a guess into an engineering decision.